How m-prunty/call_me_maybe works
Call Me Maybe is presented as a natural-language-to-function-call system, but the checked-in runtime shown here is much smaller: its entry point initializes the local-model SDK, downloads tokenizer vocabulary, and requests logits. The README describes constrained decoding, function selection, and schema validation as product capabilities, but the tree contains no corresponding implementation modules, so those appear as documented, unverified stages rather than confirmed runtime flow. The function-calling input files exist, but the sampled entry point does not read them.
Main entry points: README.md Function selection · call_me_maybe.py Program entry · __init__.py Local model SDK · function_calling_tests.json Prompt inputs
Components 10
Runtime inference
- Program entry
call_me_maybe.py - Local model SDK
llm_sdk/llm_sdk/__init__.py
Function calling
- Function definitions
data/input/functions_definition.json - Prompt inputs
data/input/function_calling_tests.json - Function selection
README.md - Constrained decoding
README.md - JSON validation
README.md - Structured call
README.md
Other components
- User
- Hugging Face Hub
Connections 9
- User → Program entry: runs
README.md - Program entry → Local model SDK: initializes
call_me_maybe.py - Local model SDK → Hugging Face Hub: downloads model files
llm_sdk/llm_sdk/__init__.py - User → Prompt inputs: provides prompts
README.md - Function definitions → Function selection: defines functions
README.md - Prompt inputs → Function selection: requests selection
README.md - Function selection → Constrained decoding: selects call
README.md - Constrained decoding → JSON validation: constrains output
README.md - JSON validation → Structured call: validates JSON
README.md